Psychophysiological, but Not Behavioral, Indicator of Working Memory Capacity Predicts Video Game Proficiency

Abstract

Visual working memory (VWM) is the ability to actively maintain visual information over short periods of time and is strongly related to global fluid intelligence and overall cognitive ability. In our study, we used two indices of visual working memory capacity: the behavioral estimate of capacity (K) and contralateral delay activity (CDA) in order to check whether training in a Real-Time Strategy (RTS) video game StarCraft II can influence the VWM capacity measured by the change detection task. We also asked a question whether individual differences in behavioral and psychophysiological indices of VWM can predict the effectiveness of video game training. Sixty-two participants (non-players) were recruited to the experiment. Participants were randomly assigned to either experimental (Variable environment), active control (Fixed environment), and passive control groups. Experimental and active control groups differed in the type of training received. Training consisted of 30 h of playing the StarCraft II game. Participants took part in two EEG sessions (pre- and post-training) during which they performed the VWM task. Our results showed that working memory capacity (K calculated according to Pashler's formula) increases after training in both experimental groups, but not in a control group. We have also found a correlation between average visual working memory capacity (calculated as K) and mean CDA amplitude no matter which group we are looking at. And, last but not least, we have found that we can predict the amount of improvement in the RTS video game by looking at the psychophysiological indices (CDA amplitude) recorded at baseline (before training), but only in the experimental group. We think that the strength of the psychophysiological indicator of VWM capacity might be a marker of the future success in video game acquisition.

Keywords: EEG; ERPs; action video games; trainings; visual working memory.

Figure 1

(A) Study design: two measurement sessions were carried out during the study (pre-training and post-training). Training included 30 h of playing in the real-time strategy game (StarCraft II), spread over 4 weeks. Training varied depending on the group. The Control group (CG) merges participants from Passive Control, who participated only in the measurement sessions and Active Control, who additionally played Heart Stone for 30 h (8 h in the laboratory and 22 h at home). (B) The visual-working memory task. Participants directed their attention to a cued hemifield (left of right, guided by an arrow at the beginning of each trial) and compared two arrays of colored squares (memory and test arrays) separated by a retention interval. The test array was either identical to the memory array (no-change condition) or differed by one color (change condition). Participants answered whether the two arrays were identical or not. (C) While all of the participants from training groups played as a Terran faction during training, the opponent’s race and strategy varied according to the training group type. Participants from the Variable Environment Group (VEG) could match three factions, from each could use one of five strategies. The faction and the strategy were randomly selected before each match for the Variable group. In the case of a Fixed Environment Group (FEG), participants always played against the Terran faction, which used an economic strategy.

Figure 2

Telemetric variables obtained from the StarCraft II (SC2) environment. While hours spent playing SC2 (upper left) allows us to confirm that both groups (VG: Variable Group; FG: Fixed Group) fulfill training assumptions, other variables indicate players’ proficiency. Barplots presenting first army unit creation (lower left) and first supply collection (lower right) give latencies in seconds. Asterisks indicate statistical significance: *p < 0.05, **p < 0.01.

Figure 3

The average K values for each set size in the two tests presented separately for each group. Lighter colors in the pair correspond to the pre-training measurement and darker to the post-training measurement. Asterisks indicate statistical significance: ^•p < 0.08, *p < 0.05, **p < 0.01.

Figure 4

(A) Grand average lateralized waveforms (contra—ipsi), averaged across P7/P8 electrodes, separately for all lateralized target distributions. For statistical analyses of CDA, the mean amplitude from 400 to 900 ms was used. (B) Mean CDA amplitude from 400 to 900 ms, separately for each load. Error bars denote standard errors of the mean, corrected with within-subjects comparisons. (C) Topography of the average activity at each electrode site from 400 to 900 ms. As values were averaged across paired electrodes, the topography is perfectly symmetrical. (D) Scatterplot of working memory capacity (K) averaged across loads and contralateral delay activity (CDA) averaged across loads.

Figure 5

In-game behaviors relationship to psychophysiological and behavioral variables. (A) Initial (from the 1st measurement point) contralateral delay activity (CDA) averaged across loads 4 and 5 predicts the overall number of played matches in the Variable Environment Group (VEG). Observed effect turned out to be insignificant in the Fixed Environment Group (FEG) only. (B) The averaged across training latency of first army unit creation predicts participants’ mean K value obtained in post-training measurement. The effect was significant in both groups.